1LIGO-G060638-00-Z

Glitch and veto studies in LIGO’s S5 search for gravitational wave

bursts

Erik KatsavounidisMIT

for the LIGO Scientific Collaboration

11th GWDAW, Potsdam, GermanyDecember 19, 2006

2LIGO-G060638-00-Z

Glitch study requirements

Low latency (seconds to hour), control-room feedback» Support real-time operations for immediate action on the instruments» Guide operators and shift workers

One day to one week feedback» Support “online” astrophysical analyses» Provide trends of the instruments’ behavior over longer strides

Month(s) feedback» Driven by data analyses groups, primarily, bursts and inspiral» Ultimate clean up of data for deep into the noise searches» Provide feedback for long-term planning and understanding of the

instruments Documentation, archiving and easy to use

» Still space for improvement!

3LIGO-G060638-00-Z

Tools for glitch study

Data Monitoring Tool (John Zweizig, Caltech)» True, real-time applications» General data-mining and watching processes at multiple levels

– Gravitational wave, auxiliary, data-acquisition specific» Science monitors, burstmon (Sergey Klimenko et al, Florida)

Electronic detector logbooks ! KleineWelle (Lindy Blackburn et al., MIT)

» Quasi-real time and offline processing BlockNormal (Shantanu Desai et al., PSU)

» Daily glitch studies Q-pipeline (Shourov Chatterji, Caltech and INFN)

» The LIGO instruments’ time-frequency microscope Burst and Inspiral event generators

» Event-driven in-depth analysis both online and offline

4LIGO-G060638-00-Z

BurstMon

SNR of loudest cluster Monitor glitch rate

» Noise non-stationarity andnon-Gaussianity

Monitor detector sensitivity

5LIGO-G060638-00-Z

Glitch studies with BlockNormal

Time-domain search for noise that doesn’t look like background noise Identify outlier events on single-instrument basis characterize them

using the ‘Event Display’ and Q-scans

seconds

seconds

6LIGO-G060638-00-Z

Glitch studies with KleineWelle

Use the Discrete Dyadic Wavelet Transform» Decompose time-series into a logarithmically-spaced time-frequency plane» Identify pixels that are unlikely to have resulted from noise fluctuations

Generate triggers with rate-based tuning of O(0.1)Hz» Provide information on the start time, stop time, frequency, number of time-

frequency pixels involved» Threshold on probability of event resulting from Gaussian noise (significance)

Analyze all gravitational-wave channels and a massive (300+) number of auxiliary channels in quasi-real time» Identify features in the data» Examine correlations with GW channel -- veto analysis» Study time-variability» Scan and classify single and multi-IFO outlier GW events

7LIGO-G060638-00-Z

Glitch rates so far in S5

Singles rates (in Hz), raw, (red), after category 2 data quality (green) and after cat-3 (blue) (DQ categories: see Laura’s talk)

(Hz)

(Hz)

(Hz)

8LIGO-G060638-00-Z

Low frequency glitches in H1/L1 during first part of S5

Plenty and loud glitches toward the end-of-lock

Trigger features

9LIGO-G060638-00-Z

Hourly glitches in LLO

Started Oct 3, 2006 and have been coming and going Attributed to BURT snapshots performed by the DAQ on an hourly

basis- mechanism not fully understood, but problem currently is not present

Excess counts on the top of the hour

Cou

nts/

min

10LIGO-G060638-00-Z

More on hourly glitches in LLO

Bursts of high significance, low frequency glitches

11LIGO-G060638-00-Z

Effects of high microseism

Increase of low frequency glitches

Low microseism at LLO (Oct 02, 2006) High microseism at LLO (Oct 15, 2006)

12LIGO-G060638-00-Z

GW-AUX correlations and vetoes Features studied in a first pass:

» Overlap as a function of trigger frequency and trigger amplitude» Formal veto analysis, i.e., study of the veto efficiency vs dead time, time-lag

analysis, use percentage» Cross-correlations

DARM_ERROR – ASI example in L1 over the first 103 days of S5

correlation time (sec)

even

ts

Veto

effi

cien

cy (%

)

GW threshold (KW)

Veto

effi

cien

cy (%

)

Deadtime (%)

KW thres=20

Veto

eff

(%)

Veto

use

(%)

Time-shift (in seconds) Time-shift (in seconds)

13LIGO-G060638-00-Z

Veto choices A collective analysis of correlations between kleinewelle

triggers from 300+ detector channels and the gravitational-wave channel

Environmental channels in LLO vs low threshold GW triggers (three distinct auxiliary channel thresholds):

Veto

effi

cien

cies

(%)

Veto usage (%)

Veto efficiencies (%) Interferometric channels are also

analyzed in the same wave after their ‘safety’ is established using hardware injections (see Muyngkee Sung’s talk)

14LIGO-G060638-00-Z

Channel-ranking principle Compare GW-auxiliary channel coincidences to expectation from background;

cast the answer in terms of Poisson probability (see poster by Erik K and Peter Shawhan)

Environmental channels in LLO vs low threshold GW triggers:

Num

ber o

f GW

vet

oed

even

ts

Veto

bac

kgro

und

even

ts (t

ime-

shift

s)

Veto significance for three distinct auxiliary channel thresholds, low (red), medium (green) and high (blue):

Background events from time-shifts Background events (Poisson)

Good understanding of the accidentals (background) in GW-auxiliary channels coincidences:

5 slope=1

slope=1

15LIGO-G060638-00-Z

_Channel_ GWT AxThr _Dur_ Deadtime Nveto Nbkg Prob bsc1accy 104 best: 104 0.100 0.000 % 9 0.00 6.9e-13 bsc2accx 104 best: 104 0.100 0.000 % 8 0.00 4.8e-11 bsc2accy 104 best: 101 0.100 0.003 % 8 0.05 1.8e-09 bsc3accx 104 best: 104 0.050 0.000 % 7 0.00 2.9e-09 bsc4accx 104 best: 104 0.100 0.000 % 9 0.00 6.9e-13 bsc4accy 104 best: 104 0.100 0.000 % 9 0.00 6.9e-13 bsc7accx 104 best: 104 0.100 0.000 % 8 0.00 4.8e-11 bsc8accy 104 best: 104 0.100 0.000 % 8 0.00 4.8e-11 ham1accz 104 best: 104 0.100 0.001 % 8 0.05 1.8e-09 ham3accx 104 best: 104 0.100 0.000 % 8 0.00 4.8e-11 ham7accx 104 best: 101 0.150 0.003 % 9 0.15 2.9e-09 ham7accz 104 best: 101 0.150 0.004 % 10 0.15 1.1e-10 ham9accx 104 best: 104 0.150 0.008 % 10 0.30 5.3e-09 iot1mic 104 best: 101 0.100 0.001 % 9 0.15 2.9e-09 isct1accx 104 best: 104 0.150 0.000 % 8 0.05 1.8e-09 isct1accy 104 best: 104 0.150 0.001 % 8 0.05 1.8e-09 isct1accz 104 best: 104 0.150 0.001 % 8 0.05 1.8e-09 isct1mic 104 best: 101 0.100 0.001 % 9 0.15 2.9e-09 isct4accy 104 best: 104 0.200 0.001 % 10 0.00 8.8e-15 isct4accz 104 best: 104 0.200 0.001 % 11 0.00 1e-16 isct7accy 104 best: 101 0.100 0.001 % 8 0.00 4.8e-11 isct7accz 104 best: 101 0.200 0.005 % 10 0.40 3.2e-08 lveaseisx 104 best: 104 0.100 0.000 % 8 0.00 4.8e-11 lveaseisy 104 best: 101 0.050 0.001 % 9 0.00 6.9e-13 lveaseisz 104 best: 104 0.100 0.000 % 8 0.00 4.8e-11 psl1accx 104 best: 101 0.100 0.007 % 17 0.10 2.4e-23 psl1accz 104 best: 101 0.100 0.016 % 13 1.60 1.7e-06

Veto choices in H1 for first 6 months of S5

Preliminary-work in progress!

Veto-yield on H1 single-instrument gravitational wave transients of ~10-21 sqrt(Hz) and above is at the 1% level for environmental channels and at the 10% level for interferometric channels

Resulting dead-times at the level of 0.5%

16LIGO-G060638-00-Z

H1-H2 coincidences

Coincidence analysis and event classification has provided evidence of events resulting from extreme power line glitches reflected all across the H1-H2 instruments

17LIGO-G060638-00-Z

H1-H2 coincidences Outlier H1-H2 vs closer to the noise floor H1-H2 events may be generated by different mechanisms

Cross correlograms in two days with extreme rates (high, top, and low, below)

seconds

H1-H2 corr -- Nov 24, 2005

H2-L1 corr -- Feb 11, 2006H1-H2 corr -- Feb 11, 2006

H2-L1 corr -- Nov 24, 2005

secondsseconds

seconds

even

ts

even

tsev

ents

even

ts

18LIGO-G060638-00-Z

Signal autocorrelationsH2 autocorr -- Nov 24, 2005H1 autocorr -- Nov 24, 2005

even

ts

even

ts

secondsseconds

seconds

19LIGO-G060638-00-Z

Summary and outlook

Significant progress -with respect to previous LIGO science runs- in following up features in the detectors

Multiple methods are identifying interesting events to be followed up

Numerous auxiliary detector channels analyzed in quasi-real detector in assisting detector monitoring and detector characterization

Rigorous tools for establishing veto criteria are maturing Bring to real-time as much as possible of the glitch work

so that to be able to support a real-time astrophysical search in the future